Anthropic releases ‘safe’ version of Claude Mythos AI model to public
Summary
Anthropic released a new version of its advanced AI model, called Fable 5, to the public with some usage restrictions for safety reasons. The company also provides a more powerful, unrestricted version called Claude Mythos 5 to select partners, especially for cybersecurity work, and is working with the U.S. government on testing AI security.Key Facts
- Fable 5 is the first public version from Anthropic’s Mythos class, the company’s most advanced AI models.
- Claude Mythos 5, an unrestricted and more powerful model, is available only to about 200 trusted organizations in over 15 countries.
- Anthropic restricts Mythos’ use on cybersecurity grounds because it can quickly find weaknesses in critical systems like banks and power grids.
- Most sensitive requests made to Fable 5 are redirected to a lower-capability model called Opus 4.8 to prevent misuse.
- The company hired experts and ran bug bounty programs to test the system’s security and prevent unlocking of restricted features.
- Anthropic’s AI tools lost Pentagon contracts after a disagreement with the Trump administration about surveillance and lethal weapons.
- Fable 5 pricing is $10 per million input tokens and $50 per million output tokens; tokens measure AI usage and can add up quickly.
- Anthropic leases a large, expensive datacenter from Elon Musk’s xAI to support its growth amid plans for an IPO (stock market debut).
Read the Full Article
This is a fact-based summary from The Actual News. Click below to read the complete story directly from the original source.